From: sandmann AT clio DOT rice DOT edu (Charles Sandmann)
Message-Id: <10109071548.AA19975@clio.rice.edu>
Subject: Re: gcc-301 difficulty
To: djgpp-workers AT delorie DOT com
Date: Fri, 7 Sep 2001 10:48:29 -0500 (CDT)
Cc: ST001906 AT HRZ1 DOT HRZ DOT TU-Darmstadt DOT De (Juan Manuel Guerrero)
In-Reply-To: <3B9899F8.17555.21743F@localhost> from "pavenis@lanet.lv" at Sep 07, 2001 09:57:12 AM
X-Mailer: ELM [version 2.5 PL2]
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: djgpp-workers AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp-workers AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

> > Here is the traceback. I will try to investigate more ASAP.
> > gcc -O2 makemake.c -o makemake.exe
> > Exiting due to signal SIGSEGV
> > General Protection Fault at eip=00240a7d
> > eax=00000197 ebx=00240b84 ecx=00000010 edx=00000000 esi=00240b84 edi=ffffffa4
> > ebp=00000fbc esp=00000fbc program=D:\CVS\LIB\GCC-LIB\DJGPP\3.01\CC1.EXE
> > cs: sel=018f  base=8295e000  limit=0049ffff
> > ds: sel=0197  base=8295e000  limit=0049ffff
> > es: sel=0197  base=8295e000  limit=0049ffff
> > fs: sel=016f  base=00015380  limit=0000ffff
> > gs: sel=01a7  base=00000000  limit=0010ffff
> > ss: sel=017b  invalid
> > App stack: [00346978..002c6978]  Exceptn stack: [002c63c0..002c4480]
> > 
> > Call frame traceback EIPs:
> >   0x00240a7d ___djgpp_exception_table+301

What OS? What version of make?

The esp/ebp are just less than 4Kb, so this looks very much like we are
on the DPMI provider's exception stack.  The SS is in the GDT not LDT
(which is probably why we can't get any information on it, also
leads me to believe it's part of the DPMI provider).  We are in the middle
of our exception code ourselves, so this looks very much like nested
exceptions (very bad news).

It could be also that our selector freeing code has run amock in an 
environment we haven't tested fully?  I think this is low probability.

The fact that it sometimes works says to me this is either an unintialized
variable or subject to external events (like a timer interrupt).

I would try building without selector freeing and see if that fixes.
Try building with the crt0 flag to zero memory.

Does it crash if you run cc1 individually (not from GCC? do gcc -v and
manually do the CC1 step)

> I myself have not got any crashes from gcc-3.0.X already for a rather long time.

I ran it quite a bit over the weekend and didn't see any problems (on NT 4.0)
which is how I noticed the missing asm block code :-)